Greco, Danilo, Gaggero, Giovanni Battista, Rahman, Arafatur and Hussain, Nazmul (2026) Cybersecurity Issues in AI Agents for Personal Devices: An Overview of Current Trends, Attack Surface and Risks. In: 2026 IEEE International Conference on Cyber Security and Resilience (CSR), 3-5 August 2026, Lisbon, Portugal.
Full text not available from this repository.
Official URL: https://doi.org/10.1109/CSR69374.2026.11759966
Abstract
Personal-device AI agents are evolving from passive assistants into systems that can browse the web, manage apps, access local resources, and act for users. This creates new cybersecurity risks by concentrating permissions, accounts, credentials, and automation in one orchestration layer. This paper surveys personal-device AI agents, examines OpenClaw as a case study, and identifies risks such as browser compromise, gateway exposure, unsafe tool execution, credential concentration, and vulnerable extensions. It also discusses user impacts, including data loss, unauthorized actions, account compromise, and hard-to-detect attacks. Our analysis relies on public documentation, vendor and third-party security reporting, and disclosed CVEs rather than on an independent penetration test. Finally, it outlines future research needs, arguing that adoption is outpacing mature security defenses and that personal-device AI agents should be treated as a distinct cybersecurity domain.
Repository Staff Only: item control page
Lists
Lists